Andreas Klöckner

Upsides of Gauss-Newton: - More robust? - Gives us "the closest" if the exact point can't be found Downside: - Conditioning penalty via the normal equations https://github.com/inducer/meshmode/blob/dd247039a027e9b75417e4fdddf8cc4c608b6622/meshmode/discretization/connection/opposite_face.py#L136-L153 cc @xywei @alexfikl...

The equivalent to https://github.com/inducer/arraycontext/pull/105 also needs to happen in meshmode's transforming pyopencl actx. cc @majosm

For the benefit of lazy eval/pytato, all our indirect accesses are predominantly expressed as indirect reads. Indirect writes are also possible. Which of the two has higher throughput on a...

Modal / nodal tracking

6

Right now, it's possible to feed modal data to things that expect nodal data and vice versa. Everything is just a bunch of numbers in a `DOFArray`. Especially if @thomasgibson's...

Two `resample_by_picking` in simple-dg TranslationUnit

2

I was just playing with (and being amazed by!) https://github.com/inducer/meshmode/pull/216.(https://github.com/inducer/meshmode/pull/216/commits/1de7f8f8fc3518ead4cd0191adf2238078636543 specifically, and https://github.com/inducer/pytato/commit/585576ac4a94c6bc3e38e5fc85ffa2fb165afea0) If I run `python simple.dg --lazy` and set a break point at https://github.com/inducer/arraycontext/blob/82b03d20849d9aac0f54cb4e9c00ba0f5908911c/arraycontext/impl/pytato/compile.py#L306, and then go ```...

See this FIXME: https://github.com/inducer/meshmode/blob/e179d357a283269b19b36af29487004d3873252a/meshmode/discretization/connection/direct.py#L126-L132 Left over from #220. x-ref: https://github.com/inducer/arraycontext/pull/29, https://github.com/inducer/meshmode/pull/220/commits/03d883ab663ba3ac1f36e84b461dfc9d0b63e88a cc @matthiasdiener @thomasgibson

This is a refreshed version of the reproducer for two minima in the Gauss-Newton-based inversion of the reference-to-global map in `_make_cross_face_batch` described in https://github.com/inducer/meshmode/pull/105#issuecomment-862003829. To reproduce it, clone this branch...

https://github.com/inducer/meshmode/blob/6e0dffc6f358c62e6be35d1016a62a1106e593a0/meshmode/interop/firedrake/connection.py#L513-L515 seems to require all fields to agree with the geometry dtype, which seems overly restrictive given that, as of the complex merge, Firedrake supports complex data. Florian Bruckner reported...

In the (relatively common) surjective case, we could skip emitting these global barriers: https://github.com/inducer/meshmode/blob/cdcc7f5e4ae883a31ac9da3629b0acd2289e2a34/meshmode/discretization/connection/direct.py#L310-L311) https://github.com/inducer/meshmode/blob/cdcc7f5e4ae883a31ac9da3629b0acd2289e2a34/meshmode/discretization/connection/direct.py#L342-L343 I noticed this while reviewing #192. cc @thomasgibson

This could avoid the forced (and possibly un-needed) modal-to-nodal conversion at the end of the connection: https://github.com/inducer/meshmode/blob/cdcc7f5e4ae883a31ac9da3629b0acd2289e2a34/meshmode/discretization/connection/projection.py#L219-L229 I *think* this is easy to do since it receives a connection as...